Editorial standards
The 5σ Standard
In particle physics, five sigma is the bar for claiming a discovery. We hold our own claims to that spirit: say exactly how sure science is — no more, no less.
Our promise
Five Sigma takes real, current physics research and makes it understandable without distorting it. Every article is written for two readers at once: a curious person with no physics background, who should never meet an unexplained word, and a physicist, who should find nothing to object to. When those two goals pull in different directions, accuracy wins and we work harder on the explanation.
How we choose sources
- Primary research first: peer-reviewed papers and official results from experimental collaborations. Preprints (papers not yet peer reviewed, usually on arXiv) are used when they are the newest evidence — and are always labelled as preprints.
- Then official institutions: CERN, NASA, ESA, NIST, the BIPM, the Nobel Foundation, observatories and experiment collaborations.
- Then reputable science journalism — for context only, never as the support for a physics claim.
We read the papers, not the press releases. Wikipedia can help us find our bearings, but it is never cited.
How we verify
- A claims ledger for every article. Each factual statement is recorded with the source that supports it and the exact place in that source — section, figure, table or page. A claim that can’t be traced to a source we have actually read doesn’t get published.
- Every number is recomputed. Probabilities, clock drifts, orbital speeds, particle fluxes: we reproduce each one in code that lives alongside the article, and the figures are drawn by that same code. When our calculation disagrees with a source, we say so and explain why.
- Every link is checked automatically — that it works, and that it points to the right paper. For journal articles we compare the DOI record’s title, authors and year with our reference; for preprints we check the arXiv record. These checks run before every release and again every week.
- Independent review. Before publication each article gets an adversarial accuracy review and a separate readability review that hunts for every undefined word and confusing sentence.
How sure are we? — evidence levels
Every article ends with a panel that labels its main claims by how strong the evidence is. These four levels are editorial judgements of the whole body of evidence. They are deliberately not sigma values or percentages.
Evidence level: Established
Confirmed many times, by different methods. Textbook physics.
Criteria Multiple independent experiments or methods agree; repeated over years; fits a tested theoretical framework; no serious unresolved anomaly in the core claim.
How we write it Stated plainly, without hedging: “Neutrinos have mass.”
Evidence level: Strong evidence
Well supported and widely accepted, but some details are still being tested.
Criteria At least one high-quality, peer-reviewed result with independent support, or several consistent indirect lines of evidence. Open questions concern details, not existence.
How we write it “The evidence strongly indicates…”, with the open detail named.
Evidence level: Hint / tension
Something doesn’t fit, or an early signal that isn’t confirmed. Could be new physics, a fluke, or an unnoticed error.
Criteria A single experiment, or methods that disagree; a modest, local-only or unreplicated signal; disputed systematic errors; a preprint; or a conflict with well-established results.
How we write it “A hint”, “measurements disagree”, “not yet confirmed” — always with what would settle it.
Evidence level: Speculative
An idea or hypothesis with no direct experimental evidence yet.
Criteria A theoretical proposal or an extrapolation beyond tested regimes, with no confirming observation. Interpretations that current experiments cannot tell apart are labelled here too, and we say so.
How we write it “Some physicists propose…”, “if … then …”, “there is no experimental evidence yet.”
Why the levels are not sigma values
A sigma value answers one narrow question: how surprising would this data be if nothing new were going on? It is not the probability that a claim is true, and it can’t see errors nobody has noticed. In 2011 the OPERA experiment reported neutrinos arriving faster than light at a significance of about six sigma; the cause turned out to be faulty equipment. A single striking result that contradicts a great deal of established physics is labelled Hint / tension however many sigma it reaches. And a claim can be Established without any single headline number, because it rests on many independent lines of evidence. When a source does report a significance, we quote it separately, with the source’s own qualifiers — for example “local” versus “global”. The full story is in What does “five sigma” mean?
Preprints, “firsts” and hype
- A preprint is named as such in the sentence itself — “in a preprint that has not yet been peer reviewed” — and badged in the reference list.
- We only call something a “first” when an authoritative source makes that claim, we attribute it, and we scope it precisely.
- We don’t use words like “breakthrough”, “revolutionary”, “proves” or “scientists baffled” in our own voice, and no headline promises more than the evidence level supports.
- When a single result conflicts with well-established physics, our default reading is an error or a statistical fluke until shown otherwise.
- Interpretations of quantum mechanics and similar ideas that make identical predictions are labelled as interpretations, never as settled fact.
How we cite
References are numbered in the order they first appear, in a style based on the one used by Physical Review. Journal articles link to their DOI, and where a free version exists on arXiv we link that too, so anyone can read the source. Every article has a “Cite this article” box with a plain-text and a BibTeX citation.
Corrections
- What counts: any factual error — a number, unit, name, attribution, date or mischaracterised finding — including in figures, interactive tools, captions and the evidence panel. Typos that change no fact are fixed without a note.
- How we correct: we fix the text and add a dated note to the article saying what changed. If the error affected the main conclusion or an evidence level, the note appears at the top of the article.
- Evidence-level changes caused by new results are updates, not corrections; they are logged with the date and the reason.
- In public: every correction is listed on the corrections page. We never silently change a fact.
- Videos: if a mistake started in one of our short videos, the article’s “What the video simplified” section says so explicitly.
- Timeliness: we aim to acknowledge reports within two working days and to publish fixes within a week.
Found a mistake? Report an error — a quoted passage and a source help most.
Independence
Five Sigma has no sponsors and runs no advertising at launch. If that changes, sponsored material will be clearly labelled and kept separate from our articles, and sponsors will have no say over what we write.
How we use AI tools
We use AI assistants, including Anthropic’s Claude, to help research sources, draft text, write the code behind our figures, and run our checks. AI makes mistakes, so it is never the final word: every claim must be traced to a source in the claims ledger, every number must be reproduced by code, and every reference is checked automatically against its official record. The editor is responsible for everything we publish.